One step backwards, two steps forwards: digitalisation and uncertainty in wind energy assessments

The complete digitalisation of wind energy workflows, from initial resource assessment to eventual repowering & life extension, gives us a chance to minimise project uncertainty. 

But we must take one step backwards first, before we can take two steps forwards. The adoption of unified data requirements that align all phases of project delivery, to ensure we don't wait for the adverse consequences of complex phenomena, but anticipate and alleviate them, means we must modify our procedures in a way that initially increases the uncertainty of our assessments in order to then reduce it. 

What do I mean? 

All uncertainty budgets are incomplete. You cannot evaluate the likelihood of the unforeseeable. We can only accommodate influences on uncertainty that we know about in our calculation. We cannot make allowances for "unknown unknowns". The trick is to ensure your budget is as complete as possible in the light of current knowledge, and includes all influences on uncertainty that it is possible to know about (I discuss "unknown unknowns" further in "How real is real (part 2)").

Uncertainty doesn't tell us how likely a particular outcome is. It tells us how much we learn when something else happens. If the uncertainty is high, we don't learn much. The errors bars are broad enough to hide unexpected eventualities. If the uncertainty is low, unexpected outcomes are not lost in the noise and teach us a lot about what we must have overlooked. We grow our knowledge base and revise our procedures accordingly to reduce the incompleteness of our budget.  

But initially uncertainty seems to increase. This is the apparent paradox of uncertainty evaluation: methodological improvements initially lead to an increase, rather than a decrease, in uncertainty, as we assimilate additional considerations that were previously overlooked into our assessments. This is the backwards step. However, it gives us the opportunity to take two steps forwards because the measures necessary to reduce that uncertainty become clear in the light of the new information that is available. 

Examples of this apparent paradox include the situation that has arisen with respect to the adoption of the 2nd edition of the power performance testing standard IEC 61400-12-1:2017. While this was being drafted it became clear that consideration of effects that were previously ignored, such as wind shear, resulted in an increase in uncertainty. This was not because uncertainty had actually increased, but because effects that had previously been neglected in uncertainty calculations under the terms of the previous edition were now receiving proper attention. In some quarters this led to a preference for the first edition. The correct response, however, is to take advantage of measurements of shear made available using, for example, lidar, to accommodate these wind conditions in the assessment and reduce uncertainty. 

We should not allow the incompleteness of our uncertainty budgets to lead us to believe uncertainty evaluations are a matter of convenience rather than necessity. The results remain grounded in reality, no matter how imperfect our methods for representing it are. 

Other examples are more dramatic. Established methods based on the capabilities of met mast mounted instruments have resulted in the load scenarios associated with complex wind shear being overlooked. Historically, measurements have not been acquired high enough to detect the incidence of, for example, atmospheric stability driven events like low level jets. As a consequence unanticipated failure rates have occurred, with torque variance outwith design envelopes leading to accelerated gear box failure. So we are not just talking about calculations that impact P90. We are also talking about issues that impact P50. 

Unforeseen situations represent project risk. The process of de-risking a project entails turning risk into uncertainty through methodological improvements so that the ways of reducing that uncertainty become clear. For example, measuring the diurnal, seasonal and directional prevalence of complex shear events across the entire rotor disc before a wind farm is built allows a more realistic evaluation of the uncertainty of related outcomes. The properly informed adoption of appropriate remedies, countermeasures and mitigation then reduces that uncertainty. Risk is reduced by turning it into uncertainty through measurement, which initially increases uncertainty, then taking action to reduce that uncertainty. 

We can now adopt a consistent methodology that begins with scanning lidar measurements before wind farms are built to validate the site specific CFD models used to predict their performance. Meaningful baselines for comparison post-construction can be established, since the same measurement methods are available throughout the project life-cycle. Digital twins, in the form of wind turbine simulations validated by operational data, can be embedded in digitised wind fields, in the form of lidar validated CFD, to predict and then track performance. 

Misaligned data requirements introduce scope for unanticipated events and “unknown unknowns” that make our uncertainty budgets more incomplete than they have to be on the basis of the evidence and understanding available to us at any given time. Proper handling of uncertainty entails minimising the scope for unanticipated events, maximising the degree to which your uncertainty budget is complete and represents realistic expectations about project outcomes. This requires alignment of data requirements through all phases of project delivery to ensure realistic expectations and useful understandings are acquired at the earliest opportunity. 

Consider the project life-cycle, represented schematically below:

The data requirements that have been fulfilled historically at each stage of project delivery have typically been based on the assumed capabilities of the available instruments rather than the ultimate project outcome that is to be achieved.

As a consequence these data requirements have been misaligned at each stage of project delivery, leading to unnecessary gaps in our knowledge, exceeding what arises due to the actual limitations of our capabilities. This introduces "unknown unknowns" that are an artefact of this misalignment, and not just a consequence of the limits of our knowledge.

Aligning our data requirements by adopting a consistent outcome-driven approach turns these "unknown unknowns" into "known unknowns" we can assimilate into our uncertainty evaluations.

This then gives us an opportunity to introduce methodological improvements that reduce these uncertainties.

Adopting a consistent representation of wind conditions that can be applied at all phases of project delivery and incorporates all available information allows for the minimisation of uncertainty. I have termed the combination of sophisticated simulations that embody our understanding of the physics with detailed measurement necessary to achieve this "eolics" in a TEDx talk on the subject in 2014. The talk can be viewed here and the transcript is available here.

Scanning lidar and site-specific lidar validated CFD makes information available so that you can take the question "what do I need to know to achieve successful delivery of my project" and propagate that question from the end of a successful project all they way to the initial stages of project delivery. 

Then the aligned data requirements made possible by instruments and methods now available and the data sets they generate can turn hindsight into foresight.